Analysis of Ant Colony Optimization on a Dynamically Changing Optical Burst Switched Network with Impairments

Abstract

The ability of an Ant Colony Optimization (ACO) algorithm to adapt on a dynamical network is considered. A previous ACO implementation which was tested on a static Optical Burst Switched (OBS) network with impairments has been altered to be simulated on a dynamic network where network links are brought online or offline. The factors affecting the adaptability of an ACO algorithm is studied and a solution to mitigate some of these factors is proposed. This paper shows that the chosen Pheromone Function is the greatest factor affecting an ACO's adaptability during a change and that other factors such as topology and magnitude of change has little to no affect on its adaptability. In an attempt to improve the ACO's adaptability during a change in its network, a sliding window Pheromone Function is proposed and tested yielding positive results.